Artificial Intelligence (AI) is reshaping retail operations, particularly in product returns, where operational costs, customer experience, fraud risk, and sustainability concerns intersect. The National Retail Federation estimated that total retail returns would reach approximately $849.9 billion in 2025, with 19.3% of online sales expected to be returned and 9% of all returns considered fraudulent[1]. At the same time, GenAI is creating new risks for retailers by enabling customers and fraudsters to fabricate receipts, alter product images, and generate synthetic evidence of product damage in refund claims, with Forter reporting that AI-enabled refund fraud is increasing and becoming harder to detect[2].

This project aims to examine the double-edged role of AI in retail return management by distinguishing two different stages in the AI transformation of retail return management. GenAI affects the front-end evidence and interaction layer, where customers may generate images, videos, receipts, explanations, and refund claims. Agentic AI affects the back-end decision and response layer, where retailers may deploy autonomous or semi-autonomous systems to monitor claims, detect anomalies, triage cases, recommend actions, and coordinate human intervention. Less is known about how retailers should manage the new risks created by GenAI-enabled return fraud, or how Agentic AI can be responsibly deployed to detect and mitigate such behaviours without damaging customer trust.

However, three important research gaps remain:

• First, existing studies mainly treat AI as an operational support tool and give limited attention to GenAI-enabled return fraud and customer manipulation.
• Second, there is insufficient understanding of how retailers can strategically combine human expertise with AI capabilities in managing returns and post-sale services.
• Third, little research has investigated how Agentic AI systems can autonomously monitor, detect, and mitigate fraudulent return behaviours while preserving customer trust and service quality.

Therefore, this research aims to develop a responsible AI framework for retail product return management by examining how GenAI reshapes return behaviour and fraud risks, and how Agentic AI can support explainable detection and decision-making with human oversight and consumer trust in the global context.

The study is expected to develop a sequential mixed-methods design combining behavioural experiments with machine-learning and explainable AI analysis of retail/e-commerce big data. This design enables the project to examine customer responses to AI-managed returns, detect high-risk return behaviours, and develop a responsible AI framework.

Artificial Intelligence (AI) is reshaping retail operations, particularly in product returns, where operational costs, customer experience, fraud risk, and sustainability concerns intersect. The National Retail Federation estimated that total retail returns would reach approximately $849.9 billion in 2025, with 19.3% of online sales expected to be returned and 9% of all returns considered fraudulent[1]. At the same time, GenAI is creating new risks for retailers by enabling customers and fraudsters to fabricate receipts, alter product images, and generate synthetic evidence of product damage in refund claims, with Forter reporting that AI-enabled refund fraud is increasing and becoming harder to detect[2].

This project aims to examine the double-edged role of AI in retail return management by distinguishing two different stages in the AI transformation of retail return management. GenAI affects the front-end evidence and interaction layer, where customers may generate images, videos, receipts, explanations, and refund claims. Agentic AI affects the back-end decision and response layer, where retailers may deploy autonomous or semi-autonomous systems to monitor claims, detect anomalies, triage cases, recommend actions, and coordinate human intervention. Less is known about how retailers should manage the new risks created by GenAI-enabled return fraud, or how Agentic AI can be responsibly deployed to detect and mitigate such behaviours without damaging customer trust.

However, three important research gaps remain:

• First, existing studies mainly treat AI as an operational support tool and give limited attention to GenAI-enabled return fraud and customer manipulation.
• Second, there is insufficient understanding of how retailers can strategically combine human expertise with AI capabilities in managing returns and post-sale services.
• Third, little research has investigated how Agentic AI systems can autonomously monitor, detect, and mitigate fraudulent return behaviours while preserving customer trust and service quality.

Therefore, this research aims to develop a responsible AI framework for retail product return management by examining how GenAI reshapes return behaviour and fraud risks, and how Agentic AI can support explainable detection and decision-making with human oversight and consumer trust in the global context.

The study is expected to develop a sequential mixed-methods design combining behavioural experiments with machine-learning and explainable AI analysis of retail/e-commerce big data. This design enables the project to examine customer responses to AI-managed returns, detect high-risk return behaviours, and develop a responsible AI framework.